An exploration of listener strategies in the lexical segmentation of hypokinetic dysarthric speech

Abstract

This investigation examined listener transcriptions of phrases produced by speakers with mild to severe hypokinetic dysarthria to examine individual strategies for the identification of word boundaries in connected speech. It was hypothesized that the most efficient listeners would use syllabic strength information to guide their parsing of the continuous acoustic stream. Six-hundred transcribed phrases (10 listeners ×60 phrases) were coded independently by two judges to identify (1) correct word transcriptions, (2) evidence of accurate lexical parsing, regardless of exact word identification, and (3) the proportions of accurate parsing of strong and weak syllable word onsets. Linear regression analysis of the group data revealed that correct segmentation of strong syllable word onsets predicted listener performance on segmenting weak syllable word onsets [R=0.805, F(1,29)=51.440, p<0.001]. Despite a wide range of listener performance on ``words correct,'' individual strategies for perceptual segmentation were evident only in the transcriptions for the most severe speaker. In this case, the poorest listeners exhibited disproportionate difficulty with the segmentation weak syllable word onsets. Results suggest that a listener's ability to use syllabic strength information in lexical parsing determines, in part, their ability to recognize word onsets. [Work supported by NIDCD, NIH.]